A thought, to which there is probably no easy answer.

I only run CPDN, for obvious reasons. So when task numbers are low, should the number of PCs running tasks be optimised for reduced energy consumption?

I've been running CPDN for some years now, and the PC is always running as much as it can handle 24/7. On mine I limit it to 6 tasks so that Photoshop has at least a chance of getting a word in edgeways!

Have noticed that the availablity of tasks has been ebbing and flowing a bit over the past months, but of course I don't know what it will be in the future.

In the meantime, should the project limit the number of PCs running tasks so that energy use is optimised (without getting into the obscure arguements of my processor is better than yours etc)? This is of course supposing that BOINC is even capable of doing such a thing.

Of course there are always drawbacks in implementing such a strategy such as machines not being available when needed and alienating users. Users dedicated to the CPDN cause would understand, but others may not.

Thoughts?

Regards, Martin

---

No doubt there is a mixture of machines out there, some used for other things and turned off, some used for other things and left on, and some that are dedicated to distributed computing 24/7.

Some BOINC projects do implement the concept of 'reliable machines', which can be used to optimise one or other variable - mopping up 'difficult' work units, for example. And that mechanism, or some other metric, could be used to control WU release when the queue is small. However, CPDN is a long way from being able to do this, though perhaps the promised BOINC server software upgrade will make these things possible. (Possible is not the same as probable!)

I've turned most of my machines off until there is a reliable stream of work, as I neither want to waste energy (though mine is 100% renewable, someone else can use it) or money. There are sufficient active machines to allow some idleness elsewhere.

In a reply to somebody else's mention of lack of WUs, it was mentioned that it isn't a steady stream, but comes more in batches, so it is more than likely that you'll see a whole bunch, and then nothing see-sawing along.

Different projects seem to use resources quite differently, and I notice that climate prediction seems to need a long time, and a lot of memory to be able to process accurately. I'm less likely wanting to shut-off my computer if I realize I'm processing a climate prediction type of WU.

Other projects seem to work fine with smaller time slices, so you are much more okay shutting-off a computer and resuming later, which really was the main idea behind BOINC in the first place ... basically, using spare CPU cycles that would otherwise be thrown away as wasted heat spinning a CPU doing nothing.

Currently, there are other pure-science projects available that might interest you if you like to keep your computers chugging-away 24/7, such as Einstein, MilkyWay, Cosmos. Some projects seem to have a constant stream of work available, while others can also similarly give it away in batches too.

Another thing about BOINC is that you have the choice of setting your computing allotment to different ratios, so that if more Climateprediction comes along, it can set the other work aside... maybe something like setting climateprediction to 96%, Einstein to 1%, Milkyway to 1%, Cosmos to 1%, Seti to 1%. ...Well, maybe not so extreme, but I think you know what I mean here.

Cheers,
Joe